基于局部小波分形特征的车牌定位研究  

Research on Location of Vehicle-license-plate Based on Wavelet Transform,Fractal Characters and Improved BP Neural Network

作　　者：贾志勇[1] 何通能[1] 王其聪[2] 李国军 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310032 [2]浙江大学信电系,浙江杭州310027 [3]浙江省公安高等专科学校公共基础部,浙江杭州310053

出　　处：《计算机仿真》2005年第9期208-210,共3页Computer Simulation

基　　金：浙江省自然科学基金项目(M603165)

摘　　要：提出了一种新的车牌定位算法。车牌定位是车牌识别系统中最为关键的处理之一,分形特征能较好地反映牌照区“粗糙”的特性。首先提取牌照图像灰度图的分形维数,再充分利用小波与分形的密切联系,以二次B-样条小波对车牌局部图像进行小波变换后,提取各个子图的分形维数。这些分形维组成的特征矢量能有效地标示车牌区域和非车牌区域,根据这些局部分形维特征,以改进的BP神经网络作为分类器,有效地实现了车牌的快速准确定位,定位时间在0.7 s左右,比同类方法更优。The paper proposes a new method of vehicle -license -plate location. Vehicle -license -plate location is one of the most important steps of license plate recognition. The character of fractal dimension reflects the roughness of the license - plate region. Extracting the fractal dimension of the gray scale image at first, then, using the close relation between wavelet and fractal to extract the fractal dimension of each sub vehicle - license - plate image after quadric B - spline wavelet transform. These characters can mark the license - plate - region and non - license- plate- region well. With these characters and BP neural network classifier, we can locate the license plate rapidly. This method surpasses other similar methods.

关 键 词：车牌定位 小波变换 分形 分形维 神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]